The Power of Negative Thinking: Exploiting Label Disagreement in the Min-cut Classification Framework
COLING (Posters)(2008)
摘要
Treating classication as seeking minimum cuts in the appropriate graph has proven ef- fective in a number of applications. The power of this approach lies in its abil- ity to incorporate label-agreement prefer- ences among pairs of instances in a prov- ably tractable way. Label disagreement preferences are another potentially rich source of information, but prior NLP work within the minimum-cut paradigm has not explicitly incorporated it. Here, we re- port on work in progress that examines several novel heuristics for incorporating such information. Our results, produced within the context of a politically-oriented sentiment-classication task, demonstrate that these heuristics allow for the addition of label-disagreement information in a way that improves classication accuracy while preserving the efcienc y guarantees of the minimum-cut framework.
更多查看译文
关键词
work in progress,minimum cut
AI 理解论文
溯源树
样例
生成溯源树,研究论文发展脉络